maxent: Estimating Probabilities via Maximum Entropy: Improved...

Description Usage Arguments Details Value Author(s) References See Also Examples


NOTE: This is a copy of the FD::maxent function, included in rexpokit to avoid the dependency on the package FD. maxent returns the probabilities that maximize the entropy conditional on a series of constraints that are linear in the features. It relies on the Improved Iterative Scaling algorithm of Della Pietra et al. (1997). It has been used to predict the relative abundances of a set of species given the trait values of each species and the community-aggregated trait values at a site (Shipley et al. 2006; Shipley 2009; Sonnier et al. 2009).


maxent(constr, states, prior, tol = 1e-07, lambda = FALSE)



vector of macroscopical constraints (e.g. community-aggregated trait values). Can also be a matrix or data frame, with constraints as columns and data sets (e.g. sites) as rows.


vector, matrix or data frame of states (columns) and their attributes (rows).


vector, matrix or data frame of prior probabilities of states (columns). Can be missing, in which case a maximally uninformative prior is assumed (i.e. uniform distribution).


tolerance threshold to determine convergence. See ‘details’ section.


Logical. Should lambda-values be returned?


This is a copy of the FD::maxent function, included in rexpokit to avoid the dependency on the package FD. Its authorship information is Authored by: Bill Shipley (; Ported to FD by Etienne Laliberte. It was copied to rexpokit by Nick Matzke (just in order to avoid the dependency on package "FD").

Having BioGeoBEARS depend on package "FD" was sometimes problematic, as it had a variety of FORTRAN code and dependencies that could slow/stall installation, particularly on older Windows machines or machines without appropriate compilers. The maxent function uses only the FORTRAN file itscale5.f, so that code was included in rexpokit, in order to include all of the FORTRAN code in a single package (greatly simplifying the compilation and code-review process for BioGeoBEARS, which is pure R.)

The function maxent is used in BioGeoBEARS only for the simple purpose of putting a probability distribution on the ordered variable "number of areas in the smaller daughter range" at cladogenesis. For example, if mx01v = 0.0001 (the DEC model default), then the smaller daughter range will have a 100 percent probability of being of size 1 area during a vicariance event (thus, the "v" in "mx01v"). If mx01v = 0.5 (the DIVALIKE model default), then the smaller daughter range will have an equal chance of being any range of size less than the parent range. If mx01y = 0.9999 (the BAYAREALIKE default), then the "smaller" daughter at sympatry (mx01y, y is sYmpatry) will have 100 percent probability of being the same size as its sister (i.e., the same range as the sister, i.e. "perfect sympatry" or "sympatry across all areas").

Original description from FD::maxent follows for completeness, but is not relevant for rexpokit/BioGeoBEARS.

The biological model of community assembly through trait-based habitat filtering (Keddy 1992) has been translated mathematically via a maximum entropy (maxent) model by Shipley et al. (2006) and Shipley (2009). A maxent model contains three components: (i) a set of possible states and their attributes, (ii) a set of macroscopic empirical constraints, and (iii) a prior probability distribution q = [qj].

In the context of community assembly, states are species, macroscopic empirical constraints are community-aggregated traits, and prior probabilities q are the relative abundances of species of the regional pool (Shipley et al. 2006, Shipley 2009). By default, these prior probabilities q are maximally uninformative (i.e. a uniform distribution), but can be specificied otherwise (Shipley 2009, Sonnier et al. 2009).

To facilitate the link between the biological model and the mathematical model, in the following description of the algorithm states are species and constraints are traits.

Note that if constr is a matrix or data frame containing several sets (rows), a maxent model is run on each individual set. In this case if prior is a vector, the same prior is used for each set. A different prior can also be specified for each set. In this case, the number of rows in prior must be equal to the number of rows in constr.

If q is not specified, set pj = 1 / S for each of the S species (i.e. a uniform distribution), where pj is the probability of species j, otherwise pj = qj.

Calulate a vector c = [ci] = {c1, c2, ..., cT}, where ci = sum(tij); i.e. each ci is the sum of the values of trait i over all species, and T is the number of traits.

Repeat for each iteration k until convergence:

1. For each trait ti (i.e. row of the constraint matrix) calculate:

gamma_i(k) = ln(t.mean_i / [sum(pj(k) tij)]) (1 / ci)

This is simply the natural log of the known community-aggregated trait value to the calculated community-aggregated trait value at this step in the iteration, given the current values of the probabilities. The whole thing is divided by the sum of the known values of the trait over all species.

2. Calculate the normalization term Z:

Z(k) = sum(pj(k) e^(gamma_i(k) tij) )

3. Calculate the new probabilities pj of each species at iteration k+1:

pj(k+1) = [pj(k) e^(gamma_i(k) tij)/ Z]

4. If |max(pj(k+1) - pj(k))| <= tolerance threshold (i.e. argument tol) then stop, else repeat steps 1 to 3.

When convergence is achieved then the resulting probabilities (pj_hat) are those that are as close as possible to qj while simultaneously maximize the entropy conditional on the community-aggregated traits. The solution to this problem is the Gibbs distribution:

Note: equation not shown in HTML help file: please refer to PDF manual.

This means that one can solve for the Langrange multipliers (i.e. weights on the traits, lamda_i) by solving the linear system of equations:

Note: equation not shown in HTML help file: please refer to PDF manual.

This system of linear equations has T+1 unknowns (the T values of lambda plus ln(Z)) and S equations. So long as the number of traits is less than S-1, this system is soluble. In fact, the solution is the well-known least squares regression: simply regress the values ln(pj_hat of each species on the trait values of each species in a multiple regression.

The intercept is the value of ln(Z) and the slopes are the values of lambda_i and these slopes (Lagrange multipliers) measure by how much the ln(pj_hat), i.e. the ln(relative abundances), changes as the value of the trait changes.

FD::maxent.test provides permutation tests for maxent models (Shipley 2010).



vector of predicted probabilities


vector of final moments


Shannon entropy of prob


number of iterations required to reach convergence


lambda-values, only returned if lambda = T


macroscopical constraints


states and their attributes


prior probabilities


Bill Shipley, original URL:

Ported to FD by Etienne Laliberte.


Della Pietra, S., V. Della Pietra, and J. Lafferty (1997) Inducing features of random fields. IEEE Transactions Pattern Analysis and Machine Intelligence 19:1-13.

Keddy, P. A. (1992) Assembly and response rules: two goals for predictive community ecology. Journal of Vegetation Science 3:157-164.

Shipley, B., D. Vile, and E. Garnier (2006) From plant traits to plant communities: a statistical mechanistic approach to biodiversity. Science 314: 812–814.

Shipley, B. (2009) From Plant Traits to Vegetation Structure: Chance and Selection in the Assembly of Ecological Communities. Cambridge University Press, Cambridge, UK. 290 pages.

Shipley, B. (2010) Inferential permutation tests for maximum entropy models in ecology. Ecology in press.

Sonnier, G., Shipley, B., and M. L. Navas. 2009. Plant traits, species pools and the prediction of relative abundance in plant communities: a maximum entropy approach. Journal of Vegetation Science in press.

See Also

FD::functcomp to compute community-aggregated traits, and FD::maxent.test for the permutation tests proposed by Shipley (2010).

Another faster version of maxent for multicore processors called maxentMC is available from Etienne Laliberte ( It's exactly the same as maxent but makes use of the multicore, doMC, and foreach packages. Because of this, maxentMC only works on POSIX-compliant OS's (essentially anything but Windows).


# an unbiased 6-sided dice, with mean = 3.5
# what is the probability associated with each side,
# given this constraint?
maxent(3.5, 1:6)

# a biased 6-sided dice, with mean = 4
maxent(4, 1:6)

nmatzke/rexpokit documentation built on July 13, 2020, 2:13 p.m.